IT / Developer Sourcing
GitHub ground-truth + people-DBs → a developer shortlist with skill proof.
Source engineers for an IT role using two parallel legs — GitHub (free, ground-truth skill proof) and the people-DBs (breadth: devs not on GitHub + employer context) — then merge, qualify against the JD, and deliver a shortlist with a rich dossier.
Why GitHub leads for code roles
The GitHub API is free (token; ~5k REST + 5k GraphQL/hr, 30 searches/min) and gives evidence a people-DB can't: real languages/repos/stars, a personal email (commit harvest), and website/LinkedIn/socials. But it only covers OSS-active devs — the people-DB leg backstops everyone else.
Dev vs. infra — pick the primary by role type.
- Code-writing roles (SWE, data/ML, backend/frontend) → GitHub is primary.
- Infrastructure / sysadmin / network / IT-support / ops / Windows-AD-M365 admin →
these people don't write public code → skip GitHub, make the people-DBs primary
(title + skills + location). Skill keywords go in
skills, not languages.
Capability chain
IT JD → PARSE (AI) → ┌─ GitHub leg: search_users → GraphQL dossier → email harvest → [Firecrawl+AI deep context]
└─ people-DB leg: people_search (aiark → …)
→ MERGE + dedup (prioritize GitHub-sourced) → QUALIFY 0–10 → shortlist (LinkedIn-first + personal email)Steps
Parse the role
Extract titles, languages/skills, seniority, location. Decide dev-vs-infra to pick the primary leg.
GitHub leg (code roles)
search_users by language + location → GraphQL dossier (repos, stars, top languages) →
harvest a personal email from commits → optional Firecrawl + AI deep context from the
dev's site.
People-DB leg
people_search waterfall by title + skills + location for breadth and employer context.
Merge, dedup, enrich the history, qualify
Dedup by LinkedIn URL, prioritize GitHub-sourced (they carry skill proof). A people-DB row shows
the current role only, so run linkedin_profile on those rows for a dated experience[] before
scoring — then score 0–10 against the JD on the dossier plus that history, and drop the no-fits.
hyreflow qualify reports whether each candidate was scored on work_history or title_only.
Deliver
LinkedIn-first shortlist with personal email and the dossier. Candidate channel only — never work email.
GitHub and clinicaltrials are free public providers — the GitHub leg costs nothing.
You only spend credits on the people-DB leg and any paid enrichment of the passes.